On Fri, May 18, 2012 at 09:09:03AM -0700, Tejun Heo wrote:

On Thu, May 17, 2012 at 10:59:51PM -0400, koverstreet-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org wrote:

quoted

@@ -729,7 +729,7 @@ static struct bio *bio_chain_clone(struct bio **old, struct bio **next,
 	}
 
 	while (old_chain && (total < len)) {
-		tmp = bio_kmalloc(gfpmask, old_chain->bi_max_vecs);
+		tmp = bio_clone_kmalloc(old_chain, gfpmask);
 		if (!tmp)
 			goto err_out;
 

@@ -751,13 +751,9 @@ static struct bio *bio_chain_clone(struct bio **old, struct bio **next,
 			if (!bp)
 				goto err_out;
 
-			__bio_clone(tmp, &bp->bio1);
-

This effectively swaps the order of bio_split() and __bio_clone().  Is
that safe?  Also, please cc the maintainer.

Now that I look at that code some more, I'm not sure it was quite right
before - that or I don't follow how it's supposed to work.

That bio_split() effectively does its own clone, so it seems to me the
bio_kmalloc()/bio_clone_kmalloc() should only be happening in the non
split case.

Also, there's a bio_chain_clone in drivers/block/osdblk.c that looks
like it's doing something similar, probably we should have some generic
code for this.